Lack of Association of Transforming Growth Factor-β1 Polymorphisms and Haplotypes with Prostate Cancer Risk in the Prostate, Lung, Colorectal, and Ovarian Trial

Introduction

Transforming growth factor β (TGF-β) plays a key role in cell cycle control and carcinogenic processes. TGF-β regulates growth and proliferation of cells and may play a dual role in carcinogenesis, inhibiting early-stage growth and promoting later-stage growth (1). TGF-β1, the predominant TGF-β, stimulates cell differentiation and inhibits epithelial cell proliferation in nonmalignant prostate cells; however, prostate cancer cells exhibit resistance to the growth-inhibitory effect of TGF-β1 (2).

Gene variants in TGFB1 have been related to functional effects. Thus, the gene would seem to be a good prostate cancer susceptibility candidate (2). The TGFB1 10Pro (C) and tightly linked −509T alleles are reported to increase levels of TGF-β1 (3-6). The −800A polymorphism is expected to reduce the affinity for the cAMP-responsive element binding protein family of transcription factors and, thus, to decrease expression of TGF-β1 (4). The Arg-to-Pro polymorphism at codon 25 was associated with lower TGF-β1 production (7, 8). The Thr-to-Ile polymorphism at codon 263 is located in the part of the TGF-β1 pro-protein that is cleaved from the active part of the protein and may thus affect TGF-B1 activation (9).

Japanese males with the TGFB1 TC (Leu/Pro) or TT (Leu/Leu) genotype at codon 10 (+29 position) were reported to have a 1.6-fold increased risk for prostate cancer (10), and American physicians were reported to have a 2.4-fold increased risk for advanced stage prostate cancer in relation to the T allele at TGFB1 position −509, whereas no excess was noted for the codon 10 variant (11).

In the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, we investigated the role of five single-nucleotide polymorphisms (SNP) in TGFB1, chosen because of potential functional significance, in relation to prostate cancer risk. We studied >1,300 prostate cancer cases and a similar number of controls.

Materials and Methods

Study Setting: The PLCO Trial

This nested case-control study was conducted within the screening arm of the PLCO Trial, which was designed to evaluate the effectiveness of prostate, lung, colorectal, and ovarian cancer screening and to investigate etiologic factors and early markers of cancer (12, 13). Participants in the PLCO Trial, ages 55 to 74 years, were recruited at 10 centers in the United States (Birmingham, AL; Denver, CO; Detroit, MI; Honolulu, HI; Marshfield, WI; Minneapolis, MN; Pittsburgh, PA; Salt Lake City, UT; St. Louis, MO; and Washington, DC) between September 1993 and June 2001.

Study Population

Men randomized to the screening arm were eligible for the nested case-control study if they had at least one valid screening for prostate cancer (prostate-specific antigen and/or digital rectal exam) before October 1, 2001 (the censor date for this analysis), completed the baseline risk factor questionnaire, provided a blood sample, and signed the informed consent for etiologic studies of cancer (n = 26,975). All men were followed from their initial valid prostate cancer screen (prostate-specific antigen and/or digital rectal exam) to first occurrence of prostate cancer, loss to follow-up, death, or October 1, 2001, whichever came first. Cases were defined as men diagnosed with adenocarcinoma of the prostate. The eligible group included 1,320 prostate cancer cases (1,213 non-Hispanic Caucasians and 107 African Americans). We selected 1,842 controls (1,433 non-Hispanic Caucasians and 409 African Americans) using risk-set sampling frequency matched by age (5-year intervals), race (whites, 1:1.2; blacks, 1:4), time since initial screening (1-year time windows), and year of blood draw.

Questionnaire Data

At enrollment, all participants were asked to complete a questionnaire including age, ethnicity, education, occupation, current and past smoking behavior, alcohol consumption, history of cancer and other diseases, use of selected drugs, recent history of screening exams, and prostate-related health factors.

Genotyping

Genotype analysis was done at the National Cancer Institute Core Genotyping Facility.⁷ All TaqMan assays (Applied Biosystems, Inc.) were optimized on the ABI 7900 HT detection system with 100% concordance with sequence analysis of 102 individuals listed on the SNP500Cancer database (14).⁸

We selected five SNPs with potential functional significance in TGFB1 for analysis: −1639G>A (rs1800468: −800G>A), −1348C>T (rs1800469: −509C>T), Ex1-327C>T (rs1982073: L10P), Ex1-282C>G (rs1800471: P25R), and Ex5-73C>T (rs1800472: T263I).

The genotype distribution of TGFB1 Ex5-73C>T (T263I) deviated from Hardy-Weinberg proportions in Caucasian controls (P = 0.05, exact test); however, the concordance rate for the quality control samples (n = 247), which were replicates from 48 study subjects interspersed throughout each batch, was 100% for TGFB1 Ex5-73C>T (T263I). Therefore, we do not believe that the slight deviation from Hardy-Weinberg proportions for this SNP is due to laboratory error.

Statistical Analysis

To estimate the risk of prostate cancer in relation to SNP genotype, odds ratios (OR) and 95% confidence intervals (95% CI) were calculated using conditional logistic regression in Caucasians and African Americans separately. The analyses were conditioned on the matching factors (i.e., age, time to diagnosis, and year of blood draw). The most prevalent homozygous genotype was used as the reference group. For each SNP, tests for trend were conducted by assigning the ordinal values 1, 2, and 3 for the homozygous wild-type, heterozygous, and homozygous variant genotypes, respectively, and by modeling these scores as a continuous variable.

Haplotype analyses were conducted using the haplo.stats package⁹ in the R program (v. 2.2.1),¹⁰ which uses an expectation-maximization algorithm to estimate haplotypes from genotype data (15). Haplotypes were estimated separately for Caucasians and African Americans excluding subjects with all the genotype data missing. A generalized linear model was used to estimate the effect of individual haplotypes by fitting an additive model for each haplotype, adjusting for the matching factors (age, time to diagnosis, and year of blood draw) in each ethnic group. The overall difference in haplotype frequencies between cases and controls was assessed using a global score test.

Results

Age distribution was similar between cases and controls; mean age was 64.8 ± 5.0 years for cases and 64.5 ± 5.0 years for controls. The proportion of African Americans was higher in controls (22.2%) than in cases (8.1%) as a result of the 4:1 matching scheme for this population. Mean body mass index varied between cases (27.2 kg/m²) and controls (27.6 kg/m²), with marginal significance (P = 0.05). Family history of prostate cancer was a significant risk factor for prostate cancer (P < 0.001). Smoking and alcohol consumption were not different between cases and controls (P = 0.78 and 0.95, respectively).

Minor allele frequencies for five SNPs of TGFB1 in control subjects ranged from 0.01 for Ex5-73T (263T) in African Americans to 0.44 for Ex1-327C (10C) in African Americans. None of the five SNPs studied was significantly associated with prostate cancer risk (Table 1 ). No dominant or recessive effect was found and no test for trend was significant (P > 0.30). Analysis of high-stage (stage ≥III) or high-grade prostate cancer (Gleason score ≥7) also showed similar results (data not shown). Results from haplotype analysis were consistent showing no significant results in Caucasians and African Americans, respectively (Table 2 ).

Discussion

Our large study suggests that selected genetic polymorphisms of potential functional significance in TGFB1 do not play a role in prostate cancer. For Caucasians, our study was sufficiently large to evaluate two previously reported associations (10, 11) with >90% power to detect an OR ≥1.5 (for dominant effect, with minor allele frequency of 0.05 and α = 0.05). In one previous study, in Asians (10), the frequency of Ex1-327C (10C) was 0.54 and the relative risk was 1.6, whereas in the other study, primarily in Caucasians (11), the frequency of −1348T (−509T) was 0.26 and the relative risk was 2.4. Although we presented data for African Americans, sample size was small and conclusions for this group are limited.

The reported association in Caucasians (11) between −1348T (−509T) and prostate cancer was largely limited to cases with extraprostatic or distant metastatic tumors (stage ≥III; n = 157), although no significant association was found for high-grade cases (Gleason score ≥7; n = 133). In our study, we found no association between the selected SNPs and high-stage (stage ≥III) or high-grade prostate cancer (Gleason score ≥7). Our study does not support the reported associations (10, 11) in Caucasians.

Our SNP selection strategy focused on variants of potential functional significance; we did not fully characterize risk in relation to all variation in this gene. HapMap¹¹ reports 21 SNPs in TGFB1, but the majority of variants are not seen in Caucasians and only six polymorphisms have a frequency ≥1% in this population group; of these six SNPs, only the T263I polymorphism was represented in our study; the remaining four SNPs in our study (−800G>A, −509C>T, L10P, and P25R) were not included in HapMap. Our haplotype-based analysis addressed potential cis relationships of the five studied SNPs, finding no further associations of interest. Full haplotype characterization would likely require more extensive genotyping; our results for potentially functional SNPs in TGFB1 suggest that this additional effort may not be warranted.

In summary, we found no evidence of association of prostate cancer with five TGFB1 genetic polymorphisms and their haplotypic combinations in PLCO trial.